Aerospace Technologies Advancements Fig Part 6 doc

Evaluation of Anomaly Detection Capability for Ground-Based Pre-Launch Shuttle Operations 163 due to realizations being based on independent experiments, and as such should still fall within the confidence bands formed in both Figs. 11 and 12. The light blue highlights superimposed over the failure injected scores in dark blue represent the “ground truth” time of failure injection and duration, so as to give a feel for the false alarm and correct detection rates. Evidently, there is a clear bifurcation between nominal and anomalous scores for both IMS and SVM, and for Orca the same is true although it is less apparent. As can be discerned from Figs. 11 – 13, we have identified the fact that both in complexity and accuracy, IMS seems to be the best choice among all of the algorithms investigated. However, there is some overlap in the confidence intervals for IMS and SVM AUC values, and the alert thresholds applied for both corresponding ROC curves yield almost identical true positive rates. 6. Conclusion and next steps We have provided a thorough end-to-end description of the process for evaluation of three different data-driven algorithms for anomaly detection. Through optimization of algorithmic parameters using the AUC, we were able to choose parameters yielding the best detection capability. The respective ROC curves corresponding to these parameters were then used to inform alert threshold selection by enforcement of a maximum allowable false alarm rate. It was found that IMS was the best performing algorithm when considering both computational complexity and accuracy. However, when evaluating the results based upon accuracy alone, the OCVSM approach is competitive with IMS due to overlapping confidence intervals present in the accuracy results. In subsequent research studies, we will provide results of unseen hold out test cases to which optimized parameters and thresholds will be applied, in order to provide additional evidence demonstrating the superiority of a particular algorithmic technique. Furthermore, we will employ a variant of the AUC that only considers performance evaluation for algorithmic comparison restricted to low false positive rates. A slightly modified definition of false alarms and missed detections that accounts for pre-defined latencies and prediction horizons will also be investigated. 7. Acknowledgements The author would like to acknowledge the support of the Ares I-X Ground Diagnostics Prototype activity, which was funded by the NASA Constellation Program, by the NASA Exploration Technology Development Program, and by the NASA Kennedy Space Center Ground Operations Project. Furthermore, the author graciously acknowledges the reviews of Dr. Mark Schwabacher, Bryan Matthews, and Dr. Ann Patterson-Hine. The author also extends appreciation to John Wallerius for his contribution of the subsection on IMS score computations, and his ideas pertaining to next steps on consideration of performance evaluation for algorithmic comparison restricted to low false positive rates. Finally, the author acknowledges the permission to use of Figs. 7 and 8 from Dr. Santanu Das, and Fig. 1a with supporting text from David Iverson. 8. References Bay, S.D. & Schwabacher, M. (2003). Mining distance-based outliers in near linear time with randomization and a simple pruning rule, Proceedings of The Ninth ACM SIGKDD Aerospace Technologies Advancements 164 International Conference on Knowledge Discovery and Data Mining, pp. 29–38, New York, NY, 2003. Cavanaugh, K. (2001). An integrated diagnostics virtual test bench for life cycle support, Proceedings of the IEEE Aerospace Conference, pp. 7–3235–7–3246, ISBN: 0-7803-6599-2, Big Sky, Montana, March 2001. Cohen, G.; Hilario, M. & Pellegrini, C. (2004). One-class support vector machines with a conformal kernel: a case study in handling class imbalance, In: Structural, Syntactic, and Statistical Pattern Recognition, A. Fred et al. (Eds.), pp. 850–858, Springer-Verlag, Berlin, Heidelberg, Germany. Das, S.; Srivastava, A. & Chattopadhyah, A. (2007). Classification of Damage Signatures in Composite Plates using One-Class SVM’s, Proceedings of the IEEE Aerospace Conference, Big Sky, MO, March 2007. Fragola, J.R. (1996). Space shuttle program risk management, Proceedings of the International Symposium on Product Quality and Integrity: Reliability and Maintainability Symposium, ISBN: 0-7803-3112-5, pp. 133–142, Jan 1996. Hart, G.F. (1990). Launch commit criteria performance trending analysis, Annual Proceedings of the Reliability and Maintainability Symposium, pp. 36–41, Jan 1990. Iverson, D.L.; Martin, R.; Schwabacher, M.; Spirkovska, L.; Taylor, W.; Mackey, R. & Castle. J.P. (2009). General purpose data-driven system monitoring for space operations, Proceedings of the AIAA Infotech@Aerospace Conference, Seattle, Washington, April 2009. Mackey, R.; James, M.; Park, H. & Zak. M. (2001). BEAM: Technology for autonomous self- analysis, Proceedings of the IEEE Aerospace Conference, Big Sky, MT, 2001. Martin, R. (2007). Unsupervised anomaly detection and diagnosis for liquid rocket engine propulsion, Proceedings of the IEEE Aerospace Conference, Big Sky, MT, March 2007. Martin, R.; Schwabacher, M.; Oza, N. & Srivastava, A. (2007). Comparison of unsupervised anomaly detection methods for systems health management using Space Shuttle main engine data, Proceedings of the 54 th Joint Army-Navy-NASA-Air Force Propulsion Meeting, Denver, CO, May 2007. Paté-Cornell, E. & Dillon, R. (2001). Probabilistic risk analysis for the NASA space shuttle: a brief history and current work. Reliability Engineering & System Safety, Vol. 74, No. 3, (Dec. 2001) pp. 345 – 352. Paté-Cornell, E. & Fischbeck, P.S. (1994). Risk Management for the Tiles of the Space Shuttle. Interfaces, Vol. 24, No. 1, (Jan. – Feb. 1994) pp. 64–86. Schwabacher, M. (2005). Machine learning for rocket propulsion health monitoring, Proceedings of the SAE World Aerospace Congress, pp. 1192–1197, Dallas, Texas, 2005. Schwabacher, M. & Waterman, R. (2008). Pre-launch diagnostics for launch vehicles, Proceedings of the IEEE Aerospace Conference, Big Sky, MT, March 2008. Schwabacher, M; Aguilar, R & Figueroa, F. Using Decision Trees to Detect and Isolate Simulated Leaks in the J-2X Rocket Engine, Proceedings of the IEEE Aerospace Conference, Big Sky, MT, 2009. Tu, H.; Allanach J.; Singh S.; Pattipati K.R. & Willett, P. (2006). Information integration via hierarchical and hybrid Bayesian networks. IEEE Transactions on Systems, Man and Cybernetics, Part A, Vol. 36, No. 1 (Jan. 2006) pp.19–33. Tumer, I. (2005). Design methods and practices for fault prevention and management in spacecraft, Technical report, NASA Ames Research Center, 2005. 9 Design Solutions for Modular Satellite Architectures Leonardo M. Reyneri, Claudio Sansoè, Claudio Passerone, Stefano Speretta, Maurizio Tranchero, Marco Borri, and Dante Del Corso Politecnico di Torino ITALY 1. Introduction The cost-effective access to space envisaged by ESA would open a wide range of new opportunities and markets, but is still many years ahead. There is still a lack of devices, circuits, systems which make possible to develop satellites, ground stations and related services at costs compatible with the budget of academic institutions and small and medium enterprises (SMEs). As soon as the development time and cost of small satellites will fall below a certain threshold (e.g. 100,000 to 500,000 €), appropriate business models will likely develop to ensure a cost-effective and pervasive access to space, and related infrastructures and services. These considerations spurred the activity described in this paper, which is aimed at: 1. proving the feasibility of low-cost satellites using COTS (Commercial Off The Shelf) devices. This is a new trend in the space industry, which is not yet fully exploited due to the belief that COTS devices are not reliable enough for this kind of applications; 2. developing a flight model of a flexible and reliable nano-satellite with less than 25,000€; 3. training students in the field of avionics space systems: the design here described is developed by a team including undergraduate students working towards their graduation work. The educational aspects include the development of specific new university courses; 4. developing expertise in the field of low-cost avionic systems, both internally (university staff) and externally (graduated students will bring their expertise in their future work activity); 5. gather and cluster expertise and resources available inside the university around a common high-tech project; 6. creating a working group composed of both University and SMEs devoted to the application of commercially available technology to space environment. The first step in this direction was the development of a small low cost nano-satellite, started in the year 2004: the name of this project was PiCPoT (Piccolo Cubo del Politecnico di Torino, Small Cube of Politecnico di Torino). The project was carried out by some departments of the Politecnico, in particular Electronics and Aerospace. The main goal of the project was to evaluate the feasibility of using COTS components in a space project in order to greatly reduce costs; the design exploited internal subsystems modularity to allow reuse and further cost reduction for future missions. Aerospace Technologies Advancements 166 Starting from the PiCPoT experience, in 2006 we began a new project called ARaMiS (Speretta et al., 2007) which is the Italian acronym for Modular Architecture for Satellites. This work describes how the architecture of the ARaMiS satellite has been obtained from the lesson learned from our former experience. Moreover we describe satellite operations, giving some details of the major subsystems. This work is composed of two parts. The first one describes the design methodology, solutions and techniques that we used to develop the PiCPoT satellite; it gives an overview of its operations, with some details of the major subsystems. Details on the specifications can also be found in (Del Corso et al., 2007; Passerone et al, 2008). The second part, indeed exploits the experience achieved during the PiCPoT development and describes a proposal for a low-cost modular architecture for satellites. 2. The PiCPoT satellite The PiCPoT design activity carried out at Dept. of Electronics, in tight cooperation with the Dept. of Aerospace Engineering and other departments of Politecnico, was aimed at developing and manufacturing a low-cost prototype of a fully operational nano-satellite. The design activity started in early 2004 and gathered about 10 people among professors and Ph.D. students, plus about 20 undergraduate students (the former for the whole Ph.D. program duration, the latter for shorter period, between 6 and 12 months each). The total effort of the project can be estimated as about 12 man-years (staff + student) for design, manufacturing and testing; a flight model and two engineering models of the PiCPoT satellite, shown in Figure 1, have been built. Fig. 1. PiCPoT engineering model. The satellite has been completely designed using COTS devices, with the exception of solar panels. The basic architecture consist of five solar panels; six battery packs; three cameras with different focal lengths; five processors in full redundancy; two RX-TX communication modules with antennas operating at 437 MHz and 2.4 GHz, respectively. The on board electronics uses six PCBs hosted in a cubic aluminum case (developed by Dept. of Aerospace Engineering), 13 cm in side and 2.5 kg total mass. The main mission was to send telemetry data (temperatures, voltages and currents) to ground, and to take, store and transmit pictures of the Earth at different spatial resolutions. The satellite was launched on July 26th 2006, from Baykonour. Unfortunately a failure of the launcher forced its destruction before being released in the planned orbit. Design Solutions for Modular Satellite Architectures 167 3. Design constraints An airborne satellite must comply with hard constraints related to the severe space environment and the inability to repair the system in case of failure. Therefore, the design and the assembly of the device must abide by tighter rules than usual good and safe design criteria applied for any electronic system. This is particularly true when using COTS components and technology, which require the adoption of design techniques which guarantee system operation even in the presence of limited faults at the device level. Other specific characteristics of a space application, although not directly related to failures of the system, further constrain the possible design solutions that can be adopted. These include the need to autonomously produce power, the limited visibility of the satellite from a ground station and the distance from it, the length of the mission, and so on. In the following, the constraints and their implications that were considered in the design of PiCPoT, along with some solutions and ideas, are outlined. 3.1 Radiation The planned orbit is close to the Van Allen belts, where a limited amount of radiation is present. This radiation might be in the form of high energy particles (protons, neutrons, alpha and beta particles) or ionizing electromagnetic rays from ultraviolet to X-rays. Due to the low orbit (polar, at 600km of altitude), and to the short lifetime assumed for the mission (3 months), total dose effects have not been considered. However, single-event effects (SEE) such as latch-up occurring in CMOS devices, and state upsets in memories and/or registers of digital circuits, might indeed induce wrong behaviors or even permanent faults. Thus the satellite circuits have been protected at the logical and system level against these events. Techniques that have been used include latch-up protection circuits, watchdog timers and redundancy at various levels. More details can be found in Sec. 4. 3.2 Electro-magnetic interference and signal integrity Noise at various frequencies may come from both internal and external sources. However, the satellite outer structure is completely metallic, and all inner circuits are therefore well shielded against electro-magnetic interference (EMI) from the outside. Internal interference between different boards or within a single board is addressed by properly designing ground planes and the printed circuit board (PCB) layout of RF and digital units. 3.3 Temperature ranges While it cycles through its orbit, the satellite alternates from broad daylight to deep Earth shadow. In these conditions, temperature may vary considerably. However, the orbital period is fast enough not to allow too much heat to build up or be released into space, preventing burning or frosting of the satellite. Thermal simulations allowed us to predict the actual temperature ranges for the outside and the inside faces of the aluminium plates that constitute the external structure of the satellite, and for the internal electronic boards. We considered the cases when the electronic boards are inactive, as well as when they are active and dissipating power (Caldera et al, 2005). The predicted outside temperature range with active electronics is [5, +50] °C; the parts subject to this range are external ones, such as solar panels and antennas. The temperature range inside the satellite is [+20, +70] °C, as shown in Figure 2, where the different curves represent the temperature of each board; all electronic circuits must comply with this range, which is compatible with standard commercial devices. Aerospace Technologies Advancements 168 Fig. 2. Thermal analysis for powered electronic boards in the satellite 3.4 Vacuum Vacuum is not a problem for sealed electronic components, but reduces the power dissipation capability due to missing convection, leaving only conduction and radiation to the outside. This problem is related to the temperature ranges outlined above. The board that dissipates more heat is the one responsible of data transmission, as it hosts the power amplifiers; we successfully tested it in a thermal vacuum chamber, with a temperature range of [-20, +50] °C and a pressure of 10Pa. While the expected pressure at the orbit altitude is some order of magnitude lower, we considered the level that we could achieve with in- house equipment sufficient to assess the board reliability. Other boards were simulated using their nominal characteristics, taking into account de-rating because of the absence of convection. 3.5 Vibrations Forces and vibrations applied to the satellite during the launch are very high, and might cause physical damages, as well as disconnection of electronic devices and disengage of electrical connectors. A careful choice of packages (i.e., no BGA devices, more sensitive to vibrations), mounting technologies and overall structure is therefore mandatory. PCBs (see Figure 3 for an example) have small size (about 12 × 8 cm 2 ), and are mechanically blocked at the four edges, therefore vibrations are kept within acceptable limits. More bulky components are secured to PCB, but connectors represent a critical point. Direct board-to- board connectors are kept in place by the mechanical fixture of boards. Other connections use flexible PCBs or small flat cables; in these cases silicon glue is used to keep in place the movable part. Specifications and requirements with respect to static loads and vibrations were established by the launcher company (Kosmotras and Yuzhnoye Design Office, http://www.yuzhnoye.com), and verified by simulations and ground tests. Mechanical tests for the maximum longitudinal g-load of 10.0g were conducted at Thales Alenia Space facilities in Torino, including random and sinusoidal vibrations. Shock and acoustic loads tests have been carried out by Yuzhnoye in Ukraine. Design Solutions for Modular Satellite Architectures 169 Fig. 3. An example of the PCB developed and used in PiCPoT 3.6 Orbit The predicted polar orbit is at a height of around 600 km (370 miles) and takes roughly 90 minutes to complete one revolution. In optimal conditions (i.e., when the satellite passes through the zenith), the line-of-sight visibility of the satellite from any given point on the Earth lasts about 10 minutes. If we take into account the distance (which varies depending on the altitude of the satellite over the horizon) and absorption due to the atmosphere, an electromagnetic signal would on average be attenuated 160 dB. Given the available power at the transmitter on the satellite, the transmitting and receiving antenna gains, and the receiver characteristics, the maximum transfer rate, assuming a certain bit error rate, can be computed. 3.7 Power The satellite has to generate its own power to function properly. The Sun is the only power source, and solar panels are used to transform light into electricity. At the Earth-to-Sun distance, the total power per square centimeter potentially available is 0.135 W. 5 out of 6 faces of the satellite are covered with solar panels, and only 3 of them are facing the Sun, with varying form factors (i.e., the angle between the solar panel and the incoming light ray). From these information, combined with orbit data the efficiency of the transformation process, the total available power can be computed. Since the satellite spends most of the time in a semi-idle state, power can be accumulated in batteries, to make it available at a later time. Our calculations show that solar panels provide an average of 1.68W of power, that we use to charge six battery packs, and gives an average power available for all electronic systems of 820mW, when worst case efficiencies of both the battery charger and the batteries themselves are taken into account. Total charge time is 63.4 hours (roughly 2.5 days), and the maximum available energy is 202kJ. Peak power consumption of the electronic subsystems can obviously exceed 820mW, provided that they are not used continuously. 3.8 Size and weight Launch costs make a considerable fraction of the total costs of a small satellite, and are directly related to the size and the weight of the satellite itself. The shape and size of the Aerospace Technologies Advancements 170 external enclosure should comply with requirements imposed by the launch vector (Kosmotras DNEPR LV, in our case), and in particular with the technique used to hold the satellite in place during launch and the way it is released when proper orbit is reached. Weight is the most important variable in computing the launch costs, since the amount of fuel needed to bring the satellite in orbit is directly proportional to it. The weight and size costs are grouped in “classes” (upper limit for weight and size); hence, the design constraint was to fit within the selected class limit, not true weight and size minimization. normal good design practice were applied in selecting components and sub-systems. 4. Design solutions Most of the design efforts for using COTS components in a satellite are aimed to protect the system from fatal events. Techniques to achieve this goal can be classified as either physical or logical. The former includes shielding the sensitive parts and choosing devices that are less prone to errors due to radiation at a comparable price tag. The latter, while allowing events to take place, mitigates or completely eliminates their effects by acting at the system level. Examples of such techniques include error correction (i.e., in memories), redundancy at several abstraction levels, and watchdog timers to reset misbehaving devices or boards. We applied several such techniques in the design of the satellite, as described in the following. Fig. 4. latch-up protection circuit 4.1 Single Event Latch-up (SEL) Latch-up (LU) occurs when a parasitic SCR made by the couple of complementary MOS devices is turned on by high input voltages (LU in ICs, caused by input over-voltages) or by high energy particles which induce a small current (this is the case for a space device) (Gray et al., 2001). The effect is a high, self-sustaining current flow, which can bring a high power dissipation and, in turn, device disruption. LU-free circuits can be designed by avoiding CMOS all-together, or by using radiation hardened devices. Since one of the goals of PiCPoT is to explore the use of COTS components for space applications, we decided to keep only some critical parts LU-free by proper device selection, and to use standard CMOS devices in other circuits, made LU-safe with specific protection circuits. The basic idea behind protection is to constantly measure current and to immediately turn the power off as soon as anomalous current consumption is detected. Once the transient event is over, normal operation can be restored. This technique is analogous to a watchdog timer, except that it actively monitors the circuit to be preserved, rather than waiting for the Design Solutions for Modular Satellite Architectures 171 expiration of a deadline. Each supply path should have its own protection circuit, which should itself be LU-free, e.g. by using only bipolar technology. The block diagram of the protection circuit of a single supply path is shown in Fig. 4, and includes: • a current sense differential amplifier (CSA), • a mono-stable circuit with threshold input, • isolating and current-steering switches (IS and CS). When the current crosses the limit set for anti-latch-up intervention (usually 2× the maximum regular current), the mono-stable is triggered and isolates the load from the power sources for about 100 ms. To fully extinguish the LU, the shunt switch (CS) steers residual current away from the load. The main problem in the design of LU protection is to balance the LU current threshold with current limit of the power supply. Namely, if the regulator current limit is activated before the LU, the current is limited but not brought to 0, and LU continues for indefinite time. 4.2 Single Event Upset (SEU) PiCPoT contains several digital circuits, including 5 processors, different kind of memories and programmable logic devices. When a high-energy particle hits a circuit, it may cause a transient change in voltage levels. While this is usually not considered a problem with analog circuits, it might adversely affect digital circuits which typically involve high speed signals with steep edges, and especially memories that rely on tiny voltages to carry their information. If the final effect results only in a glitch (Single Event Transient, SET), then it can safely be ignored; however, if the event is latched, or directly upsets a bit (or multiple bits) in a memory or a register, it will probably lead to incorrect behaviors (soft errors). In extreme cases, such as when a configuration bit of a programmable logic device turns an input into an output, it can even cause severe damages. In the less dramatic case of a soft event, we distinguished between three different kind of errors: 1. errors on dynamic data and/or in code segments resident in volatile memory; 2. errors on data stored in non-volatile memory; 3. errors on program code stored in non-volatile memory. The outcome of such events may be wrong data, wrong behavior (if the event affects some data dependent control, for instance) or even a crash (i.e., if the upset results in a non- existent op-code for a processor). The available solutions to address the problem are very diverse, each with its own advantages and shortcomings. Some cope with all three kind of errors, others address only some of them. We applied different techniques in various parts of the satellite, depending on the kind of protection we wanted to provide. The selection was driven by the need to keep the design simple and power consumption and total budget low. We did not employed radiation-hardened devices (which are too expensive and against the whole philosophy of the project to use COTS components), and memories with error corrections code (ECC, which are only useful for dynamic data and do not protect against multiple bits upsets). The susceptibility of COTS components to radiation can be very different. Careful selection of the best devices for the application allows us to strongly reduce the probability of single event upsets. We examined several kind of memories in search for the best ones, and in particular we considered: [...]... mechanical purposes (Fig 9) The inner part of the satellite is mostly left empty, to be filled by the user-defined payload This last is the only part to be designed and manufactured ad-hoc for each mission; thanks to modularity and reuse, each tile is designed only once, but manufactured and tested in relatively large quantities 180 Aerospace Technologies Advancements (a) (b) Fig 9 Example of modular... is converted into a standard ITU-R BT .65 6-4 digital stream, then the interlaced raw image is converted into a compressed JPEG picture, which is divided into 9 zones and individually sent to ground An Analog 178 Aerospace Technologies Advancements Devices Blackfin DSP manages the board and implements the compression algorithm and permanent storage of the pictures 5 .6 RF transceivers The satellite operates... indirect narrow-pitch 140 pin connectors, and a second 60 pin connector for selected signals (a) (b) Fig 5 CAD model of wirings: RF and batteries connections (a) and stackable connections among boards (b) On the remaining signals (especially for power lines, and RF connections), instead, we have to use special media: 1 76 Aerospace Technologies Advancements • SMA and coaxial cables for RF, in order... 2005) Fig 6 Nanokhod dual-track system (Image Courtesy: Klinker, 2007) 1 96 Aerospace Technologies Advancements The tracker consists of two “caterpillar” track units, a tether unit, and a payload cabin (Fig 7) The caterpillar tracks are driven by four internal drive units The drive units consist of a stepper motor attached to a 64 :1 planetary gear in front of a clown and pinion stage The output stage... connection data rate is 9.6kbps, while the SHF rate can be set from 10kbps to 1Mbps, but analyzing the link budget it is possible to show that only using the lower speeds a reliable communication is attained Some of the foreseen applications (for example high resolution or real-time imaging) require a large bandwidth downlink In this case the performances of 1 86 Aerospace Technologies Advancements the standard... independent communication links or human-operated (Table 1) Over the years, technology advancements have been made that resulted in development of new mobility systems, due 190 Aerospace Technologies Advancements to ever-growing science requirements Several such robot mobility systems have been reported in literature (Fig 1) Here, we try to systematically classify all these systems depending on the type... et al., 20 06) The rocker-bogie system is geometrically scaled to develop rovers of different mass class Until now, JPL has used it in the Rocky7 testbed, Mars Pathfinder mission rover Sojourner, Mars Exploration Rovers (MER) Spirit (Fig 4) and Opportunity It is also being used to develop the 800 kg Curiosity (MSL) rover (Fig 5) slated to be launched in 20 16 (NASA/JPL/Caltech website, 2005) Fig 4 Flight... data collisions Every data packet contains as the first part the univocal master ID If a collision occurs, all the masters involved wait for a random period before starting a new bus access 184 Aerospace Technologies Advancements 9 On-Board Computing In ARaMiS the On-Board Computing (OBC) unit is mainly responsible of managing the system, in particular of: creating and transmitting (by Transceiver... communication channels using separate antennas, at frequency of 437 MHz and 2.4 GHz respectively More details about the implementation can be found in Section 5 .6 The only non replicated unit is the camera control board (payload) 174 Aerospace Technologies Advancements 4.4 Shielding In a satellite two kind of EMI must be handled: radio-frequency interferences and radiation We developed special solutions to... developing appropriate testing strategies aimed at cheap satellites, as part of the ARaMiS project On the other hand, since each module contains a micro-controller, automatic test functions (e.g., BIST) can be embedded without any extra hardware, in order to simplify verification of correct satellite operation 188 Aerospace Technologies Advancements 13 Conclusions The paper presents the design issues of . reuse and further cost reduction for future missions. Aerospace Technologies Advancements 166 Starting from the PiCPoT experience, in 20 06 we began a new project called ARaMiS (Speretta et. with this range, which is compatible with standard commercial devices. Aerospace Technologies Advancements 168 Fig. 2. Thermal analysis for powered electronic boards in the satellite 3.4. randomization and a simple pruning rule, Proceedings of The Ninth ACM SIGKDD Aerospace Technologies Advancements 164 International Conference on Knowledge Discovery and Data Mining, pp. 29–38,